Circuit Cellar Magazine, Issue #8 (April May 1989)

Ctrl

The Whole Story

EDITORk ’ N K

S

oftware is one of the major issues we wrestle with here at Circuit Cellar INK. On the one hand, our readers look to us for practical hardware solutions and innovative control techniques. There are several other hi h-quality magazines where readers can turn if their interests center on code rather than solder. Given that wea ave to be discriminating in our selection of articles, why not focus solely on hardware? On the other hand, our subtitle reads The Computer Applications ]ournal, not The Computer and Controller Hardware Journal. We have romised practical, complete solutions, not black-box instructions. Since our readers are interested in functional appPications, shouldn’t we provide them with all the pieces required to build those applications? When you look at our table of contents for this issue, you’ll see articles on software. Yes, hardware is what makes us unique, but it’s not enough. Programmable microcontrollers and complex designs have become so common that even the most diehard hardware engineers must recognize the importance of software. It’s become fashionable to talk about how huge the programs running on embedded controllers and other applications have become. This is used as a way of saying that software is the sole force driving the development of computer applications. I’m not sure that I believe all the numbers being thrown about, but I do see signs of a growing partnership between “engineers” and “programmers.” Until recently, it was fashionable for hardware designers to disparage the work of programmers, and vice versa. Now, cooperation is common. What’s more, hardware and software engineers are learning one another’s disciplines, and finding that it makes their own work more creative and productive. Circuit Cellar INK is dedicated to helpin you become a better computer applications designer, engineer, or programmer. If we can teach you more about tR e discipline you know least, we’re well on our way. Speaking of Trends.. ~ When you get as many pressreleasesas we do hereat Circuit Cellar INK, you can, if you squint just right, see trends start to develop. One of the trends that seems to be picking up steam is the use of microprocessors rather than microcontrollers in embedded and control applications. Members of the 8088 and 68000 families are showing up in places that, until recently, were reserved for 8052s and 6809s. What’s up? For starters, it’s easy to see that hardware engineers aren’t the ones driving this trend. After all, who wants to have to work extra glue, memory, and I/O circuitry into a design if you don’t have to? It’s not as though controllers aren’t powerful enough. Controllers like the Intel 80960 are, arguably, more potent than anything the company offers on the microprocessor side. That argument aside, most control and embedded applications just don’t need the wide data path, huge address space, and 25+ MHz clock speed of the latest microprocessors. No, the driving reason for this trend can be found back about four paragraphs. Programmers and engineers who have to write the software for applications are getting more vocal about the need for better programming tools. In general, if you have an IBM PC on your desk, ou have the basis for a high-powered 8088 software development system. If you have a 68000-based desktop mat ii ine, you’re only so far from developing software for a 68000 control project. The fact is, software development time is now more valuable than mere hardware. This explains the appearance of PC-clones on just about every bus; a 4” x 6” card with 80386,l meg of memory, and VGA built-in; and 68000 UNIX single-board computers. I don’t expect to see microcontrollers and their development tools disappear any time soon. Indications are that 8-bit microcontrollers will be the major portion of the market for years to come. But where microcontrollers were the only solutions available, now you can pick from a wide variety of platforms based on the many criteria (including financial and time) of a given project. Time Marches Cz

Those with eagle eyes might have noticed than we jumped from the January/February date of Issue #7 to the April/Maydateof thisissue. Isthisasignofcorporatememoryloss,orapersonalvendettaagainstMarch? It’sneither. It is, instead, the easiest way we could find of making the monthson the cover and the real schedule match. You should notice no change in the way you’ve been getting Circuit Cellar INK.

Curtis Franklin, Jr. Editor-in-Chief April/May I 989

1

FOUNDER/ EDITORIAL DIRECTOR Steve Ciarcia PUBLISHER

Daniel Rodrigues

EDITOR-in-CHIEF

Switching Power Supplies

Curtis Franklin, Jr.

Efficient Power for Embedded Control Systems

by Steve Ciarcia

ASSOCIATE PUBLISHER

John Hayes

When you downsize the controller, the power supply has to shrink, too. Steve looks at the “black art‘ of designing efficient, clean power supplies in compact form factors.

ENGINEERING STAFF

Ken Davidson Jeff Bachiochi Edward Nisle y

CONTRIBUTING EDITOR

Thomas Cantrell

CONSULTING EDITORS

22 cl

Mark Dahmke Larry loeb

CIRCULATION COORDINATOR

Rose Mansella

Product Reviews-

The Next Generation Circuit Cellar INK sells out and enjoys if! Supercharged Worry Munchers Circuit Cellar INK looks at four diverse applications solutio

30 Writing A Real-Time Operating System-Part 2 cl Memory Management and Applications for the HD64180 by Jack Ganssle

CIRCULATION CONSULTANT

Jack Ganssle discusses working with the HD64180’s internal MMU and tips on writing applications in the conclusion of this two-part article.

Gregory Spiizfaden

PRODUCTION MANAGER

Tricia Dziedzinski

BUSINESS MANAGER

Jeannetfe Walters

Editor’s INK The Whole Story

STAFF RESEARCHERS

by Curtis Franklin, Jr.

Northeast

Eric Albert William Curie w Richard Sawyer

Reader’s INK--Leffers fo the Editor

Robert Seek Midwest

Jon Elson lim McDonough

Visible INK--Levers to the INK Research Staff

5 20

West Coast

Frank Kuechmann Mark Voorhees

From the Bench

Creating a Network-based Embedded Controller

46

by Jeff Bachiochi

Cover Illustration by Robert Tinney 2

C//?CU/T CEL LA I? INK

1

Circuit Cellar BBS-24 Hrs. 300/ 1200/2400 bps, 8bits. no parity. 1 stop bit, (203) 871-1988.

THE COMPUTER APPLICATIONS JOURNAL

ImageWise/PC-The Digitizing Continues- Part 3 Topping it off with Software

by Ed Nisley

Although solder is the favorite programming tool of many engineers, software is the glue that holds ImageWise/PC together, as this three-part series concludes, Ed Nisley describes the software that makes this PC-BUS digitizer possilbe.

rROL 1 I N T E R R U P T 1

1521 HD647180X-A New 8-Bit Microcontroller -

I:pu

Embedded Controllers Get Respect

by Tom Cantrell

PORTh

6-7

Firmware Furnace

While the state-of-the-art marches toward 32 bits, a-bit microcontrollers keep gelling more powerful. The HD647180X is the latest integrated controller built on the foundation of the proven Z80.

The True Secrets of Working with LCDs

56

by Ed Nisley

Advertiser’s Index

65

ConnecTime - Excerpts from the Circuit Cellar BBS

67

Conducted by Ken Davidson

Steve’s Own INK

Smile When You Call Me That by Steve Ciarcia

72

The schematics provided in Circuit Cellar INK are drawn using Schema from Omotion Inc. All programs and schematics in Circuit Cellar INK have been carefully reviewed to ensure that their performance is in accordance with the specifications described, and programs are posted o&the Circuit Cellar BBS for electronic transfer by subscribers. Circuit Cellar INK makes no warranties and assumes no responsibility or IiabilitY of any kind for errors in these programs or schematics or for the consequences of any such errors. Furthermore, because of the possible variation in the quality and condition of materialsand workmanship of readerassembled projects, Circuit Cellar INK disclaims any responsiblity for the safe and proper function of reader-assembled projects based upon or from plans, descriptions, or information published in Circuit Cellar INK. CIRCUIT CELLAR INK (ISSN 0896-8985) is published bimonthly by Circuit Cellar Incorporoted4Park Street.Suite2O.Vernon.CT 06066 ( 2 0 3 ) 875-275 1. Second-class postage paid at Vernon, CT and additional offices. Onevear (6 issues) subscription ;ate U.S.A. and posse&ions S14.95.CanadoS17.95.all &her dountries $26.95.. All subscription orders payable in U.S. funds only, via international postal money order or check drawn on U.S. bank. Dlrect subscription orders to Circuit Cellar INK,Subscriptions, P.O. Box 2099, Mahopac, NY 10541 or call (203) 875-2 199. POSTMASTER: Please send address changes to Circuit Cellar INK. Circulation Dept., P.O. Box 2099, Mahopac, NY 10541. Entire contents copyright 1988 by Circuit Cellar Incorporated. All rights reserved. Reproduction of this publication in whole or in part without written consent from Circuit Cellar Inc. is prohibited.

April/May I989

3

CM

READER’S ’ N K I had a few comments on the article on multitasking (“Writing a Real-Time Operating System”) and the discussion on control networks (“ConnecTime”) which appeared in issue #7of Circuit Cellar INK. As part of my own multitasking system (wearing many hats in a small company), I do hardware and software design, write our manuals, and do a monthly article for Radio World Newspaper (a technical newspaper for radio broadcast stations).

TASKO:

TASKLOOP: JSR JSR JSR JSR BRA

TASK0 ;Go do task TASK1 ;Go do task TASK2 ;Go do task TASK3 ;Go do task TASKLCCP ;Go do it

0 1 2 3 all again

We set up an area of RAM to hold the “program counters”: TASKOPC: TASKlPC: TASK2PC: TASK3PC:

DS DS DS DS

2 2 2 2

Task entry looks like this:

, -Set aside 2 bytes of ;RAM for task 0 PC

LDX TASKOPC;Get our program counter JMP 0,X ; and go to it

Task exit looks like this: TASKOF:

Multitasking The multitasking system discussed in the article looks quite complete, and complicated. We’ve been using a simpler system on 6802-based systems. Hardware interrupts are used just for I/O buffering (rather than task switching). The interrupt-driven I/O buffering (using circular buffers) allows us to send or receive blocks of I/O from different devices or users. The system uses roundrobin task switching with no priorities. Every task waits its turn. A task switch is initiated whenever the current task runs out of work. This is generally I/Orelated. Either the task is waiting for input and the input buffer is empty, or it has output and the output buffer is full. On exit, each task resets its “program counter” to the point where the decision was made that caused the exit. If there is a particularly processor-intensive task that we do not want to hog the system, it can be broken into parts. On exit, the task program counter is set to where the task is to continue. In 6800 assembly, the task manager looks like this:

Letters to the Editor

TASKOG:

-Specify buffer 0 LDAS x0 JSR BDFFULLiGo see if it's full EQ TASKOG ;Cont routine if not full LDX #TASKOF;Point where come back STX TASKOPC;Store as PC for this task ; and exit RTS ;Buffer not full, so output LDABXO ; Specify buffer 0 ;Geta bell LDAA Xl JSR PUTBUF ; and put it in buffer 0

On entry to task 0, we pick up the program counter for this task, then start executing at that point (in the example, TASKOF). We redo the test that caused the task exit (checking to see if buffer 0 is full). If it’s still full, we set up our task program counter for next time. If it!s not full, we continue with the task (here outputting a bell character). The key to this approach is that whenever a task needs to wait, it goes on to another task. When each other task is carried as far as it can go (without waiting), we come back around to see if this task is ready for action. This approach uses only the standard hardware stack. There are not separate stacks for each task. Note, however, that this complicates the use of local variables. Since many high-level languages keep the local variables on the stack (throwing them out on procedure exit), the stack would quickly get confused. The simplest approach to this problem is to just use global variables that are always allocated. Each task needs its own set of variables. If each task is running the same program (thus having the same variable requirements), each variable could become an array that specifies which task the variable is used by. To save memory, an array called Stack (Task) could be set up. Temporary variables could be stored on this “high-levellanguage” stack that is dedicated to this task. Since they are no longer needed, the stack pointer would be moved to delete them. This “high-level-language” stack allows reuse of memory for temporary variables. Of course, one purpose of the hardware stack is keeping track of subroutine return addresses. If a task gets Apr/l/May I989

5

three subroutine calls deep and then exits, all those return addresses are still on the stack. The stack will very quickly become confused again. For simplicity, all task exits would have to be from the “main-line” code. This can be accomplished byeithercheckingforanexit-causingcondition prior to calling the procedures (see if the buffer is full prior to calling Buf Put) or having a flag (ProcFail) that is set if the procedure failed and needs to be run again (and cleared if the procedure did not fail). When we get back to the main-line code, we continue if ~rocFai1 is false. If ProcFai.1 is true, we set up our task program counter to repeat the procedure next time, then exit this task. Networking

for the data. I followed this with a checksum byte, which is the sum of all bytes in the message (excluding the AA55 hex flag, but including the addresses, byte count, and data). Another approach is to use the 2’s complement of the sum as the checksum. The receiver then adds up all the bytes in the message, including the checksum. If the result is 0, the message is good. Mark had an ETX byte to mark the end of the message. Since the message already includes a packet length byte, we already know where the end of the message is, making ETX unnecessary. Contention Avoidance

Mark’s system used a fiber-optic ring, where it is not possible to get contention (multiple devices transmitting on the same medium simultaneously). Each fiber link has only one transmitter and one receiver, so a “token passing” protocol does not appear necessary. Any device wanting to send a message may transmit it at any time. The next device in the ring receives that message. If its address matches the ToAddress of the message, it acts on the message. If the address does not match, it passes the STX, Address, Command, Packet Length, data bytes, ETX complete message on to the next device in the ring. Eventually, the message should find a home. I’d suggest the following format: Note that this ring could be expanded to a multibranHex AA, Hex 55, ToAddress, FromAddress, PacketLength, ched tree or matrix where each device has several links to data bytes, CheckSum several other devices. Received messages that are not for this device are passed on to other devices through any one The AA55 hex is a beginning-of-message flag. We of several output ports. The decision as to which port to need to have a flag that will never occur in our data. Syn- use canbe based on a table held in that device. The routing chronous protocols do this by sending more than the information could also be included in the message header. maximum number of allowed 1 bits in a row in a character. The device originating the message would include several To prevent a data byte from having too many Is in a row, bytes of ToAddress. For the destination site to answer, it they use “bit stuffing” to stuff a 0 in the character after the is necessary to include several bytes for the FromAddress. maximum number of Is. On the receiving end, after the If there really were a possibility of contention (a bus maximum allowed number of Is is received, the following topology where more than one device can transmit on the “stuffed” 0 is removed, restoring the data to original. same medium simultaneously), a contention-avoidance In asynchronous communications, we cannot stuff system is desirable (though not required). If each device more bits in a character (although some protocols send a transmits at random times and the system loading is low, break as a begin-of-message flag, which is detected at the there is a good possibility the message would get through. receiver as a framing error). What we’ve done is “byte The next approach to contention avoidance would be stuffing,” quite similar to bit stuffing. If the data contains to “look before you leap.” Listen for someone else transan AA hex, we stuff a 00 after it on transmission and mitting before you bring up your carrier. This eliminates remove it on reception. This guarantees that the flag some contention, but there is a possibility of several sites (AA55 hex) will never occur in the data string. By the way, holding off until a carrier drops, then having them all come AA55 hex is used since it is an alternating pattern of OS and up at once. To avoid this problem, a delay can be introIs, then the reverse pattern. Both Mark Lampkin’s proto- duced after detecting carrier drop and before bringing up col and mine include the ToAddress. I’ve included a carrier. If this delay is fixed (and different1 for each site, a FromAddress to allow theother device to respond (whether priority system is established. The site with the shortest with an acknowledgement or with requested data). Mark delay gets in first. If a random delay is used, all sites have included a “command byte.” I believe this doesnot belong equal access. in the header, but should instead be in the data bytes area of the message. This allows for multibyte commands and Token Passing makes even “noncommand” data packets the same. If needed, the first byte of the data bytes section could be a In token passing, a “permission-to-transmit” token is “packet type” byte, which determines how the remainder passed from device to device in the system. When a device of the packet is to be interpreted. As Mark, I then left room receives the token, it transmits any messages it is holding,

In ConnecTime, there was quite a discussion of network protocols. I had a few comments (again, based on systems we manufacture). Mark Lampkin gave a suggested data packet format. His simple format did not contain a return address or error checking (he’s gotten no errors). His packet consisted of:

6

C/KU/T CELLAR INK

then transmits the token to the next device in the system. There is danger, however, of the token getting lost, due to a transmission error. When this occurs, the system must detect it and generate a new token. Absence-of-Data Token Passing In one of our systems, we’ve used “absence-of-data token passing.” Each device in the system includes a timer that is set to a “sitedelay”on reception of a valid data byte. If the timer reaches 0 before a valid byte is received, a “site counter” is incremented and the site delay timer is set back to the initial site delay. This continues until the highest site number in the system is reached. After that, the site counter is reset. When the site counter is incremented, each device in the system checks to see if the site counter now matches its site number (“it’s now my turn to talk”). If so, carrier is brought up and FE hex characters are sent for site delay and act as a ‘leader.” Once the “leader” is complete, we transmit all the data packets that are held in the transmit buffer (with no more leader). When the transmit buffer is empty, we leave the carrier idle for a couple character times to allow the last byte to get through, then drop the carrier. A SiteDelay after the last valid byte, all sites increment their SiteCounters, enabling the next site in the system. If any site does not haveanydata tosend,itleavesitscarrieroff,automatitally passing the “permission-to-transmit” token to the

next site in the system. In addition, each time a site receives avalidmessage,itsetsitssitecounter tomatch theFromAddress of the received packet. This resynchronizes all sites in the system. This “absence-of-data” token-passing system is really just like a standard token-passing system except for the form of the token. Further, no site must regenerate a lost token, since the token is the lack of data rather than the presence of a certain data sequence. It’s hard to lose something that was never there! Conclusions There are lots of different ways of handling networks. Right now, I like the absence-of-data token passing for bus systems and the routed matrix for nonbus systems. The routed matrix has some additional delay due to message retransmission, but this retransmission removes noise and timing inaccuracies from messages, resulting in better system performance. The routed matrix also has a higher throughput since the medium connecting two devices or sites is only carrying traffic that needs to go between those sites. In a bus system, all media in the system carries all the traffic. The routed matrix can be carrying several messages simultaneously over various portions of the medium. The bus can only carry one message at a time. Harold Hallikainen San Luis Obispo, CA

Still Blast ROMs?

Develo and test RBM cot ie in minutes without leaving ywyT keyboard - with The ROMul atop. ’ \

_

most f ROMulator system need not be l o c a t e d neer and Mats

GIVE A NEW DIMENSION TO PERSONAL COMPUTING .The amung V&e Master Key System adds recognition to iust about any program 01 apphcation Voice command up to 2% keyboard macros horn wilhm CAD. desktop publishing. word prccessmg, spread sheet, 01 game programs. Fully TSR and occupies less than MK. Instant response time and high recognlbon accuracy Voice recoQnl_ tion to&box utilittes are Included A genulnc pmductlvlty #n*nc#r! SPEECH RECORDING SOFTWARE Dlgtlally record your own speech. sound. 01 music to put into YOU, own sottware oloorann Software ofov~des samolino fate vat~abonr. araohlcs-based edlbna. and

1

zE?%k be 68Oi0, etc.’

ter Key Syrtsm conststs

01 P plug-m card, durable lhghtwelghl microphone headset. sohware. and manual Card Ills any available slol Exiernal ports consist 01 m~c Inputs and volume controlled output sockets Hlph qoallty throughout. uly and

Source code for host software is included

Ask us about faster (than 150ns)

ROMulators and custom cables for most non-27xxx ROM sizes.

FOR MORE INFORMATION, PLEASE CALL OR WRITE

Gmmmar

-E ELI

VIsh .n* MurcrCud hccspre* D**er I”q”me* we,comc

Gremsnar Engine, hc

ONLY $149.95 COMPLETE

ONLY $89.95 FOR TANDY lm, StJ-TL MOOELS SOFTWARE ONLY-REOUIAES TANDY BRAND ELECTRET MICROPHONE.

ORDER HOTLINE: (SW) 342-17Il Monday-Friday. SAM to 5 PM Pacihc Ttme Visa/MasterCard. company checks. money orders. CODS (with proor approval) accepted Personal checkssubject to 3 week shipping delay. Specify computer type and disk format (3%” 0~ 5%“) when ordering. Add $5 shipping charge lot delivery m USA and Canada Forelgn tnquirles contact Covox lot C 8 F ouotes 30 DAY MONEY BACK GUARANTEE IF NOT COMPLETELY SATISFIED. ONE Y E A R ~~ARRANTY ON H A RD W AR E . CALL OR WRITE FOR FREE PRODUCT CATALOG

3314 Morse Road

Columbus, Ohio 43231 614/471-1113

In Pacific and Mountain tmezones: 415/595-2252

Circle No. 120 on Reader Service Card

Circle No. 113 on Reader Service Card

Switching Power Supplies Efficient Power for Embedded Control Systems by Steve Ciarcia

W

ith all the emphasis on em- based dedicated controllers. [Editor’s bedded controllers and proc- Note: For details on the RTC31 and ess control in the pages of RTC52 controllers, see “From the Bench” Circuit Cellar INK, it was only a mat- beginning on page46.1 SO what’s masoter of time before I felt obligated to chistic about this? Everything. address the power requirements of At the same time I decided to resuch systems. After all, it hardly makes vamp the control system I thought I sense to build the world’s most effi- would take a more comprehensive cient micro-miniature controller only approach toward power control and to have it powered by the world’s distribution. My present system has most inefficient, grossly oversized, Ne- separate AC power supplies in each of anderthal-technology power supply. its control areas. While I do have At least, that’s the way my logic went. costly AC UPS power-backup units Little did I realize what masochism I oncritical computercontrolelements, was inviting. the system is still susceptible to conThe Circuit Cellar has a variety of trol errors when individual locally control systems. There is such a maze, powered sensors lose power (usually in fact, that I’m even beginning to false negative because they are no need a road map to figure out where longer operable). Of course, there are everything is going. To alleviate some various methods of redundancy and of the confusion and provide a sub- monitoring that I could incorporate to stantial development base for future correct these problems, but it seemed expansion, I will be converting much like too much bother considering the of my system to an RS-485 net with frequency of such events. Neverthemany locally intelligent data acquisi- less, I determined I’d fix thiscondition tion and control nodes. Rather than if I ever redesigned the system. stringing a wire 200 feet out to the Instead of 115VAC power for the garage to monitor a switch closure, I new controllers, I intend to run everywill simply connect it to the garage thing on +12 volts. Like commercial controller node with the dozen or so alarm systems that run battery power other I/O contacts and its status will even to the remote sensors, my new be transmitted with everything else. system could be entirely battery oper(This whole system will be described ated in the event of an AC power in Circuit Cellar INK in the coming failure. One AC-line 1Zvolt power months.) supply with a constantly charged 12Not only does this cut down on volt gel-cell (or car battery for that wiring (you wouldn’f believe how much matter) could serve as an uninterwire there is around this place already!) ruptable power source for the entire but it allows control changes to be system. The same power supply dedone in a more orderly fashion. At the sign could also be applied to a single very least I’ll have less aggravation embedded control system to provide tracing shorter wires. Most of the new UPS operation. system will consist of 8031- and 8052Running things from 12 volts is

10

Cli?CU/T ClwAR INK

Photo l- 78SObased design using traditional techniques. nothing new. This application, however, presents some very special design considerations. Individual net controllers will have a variety of tasks. Some will be simple contact closure or temperature monitors. Others will have more elaborate configurations with event printers, modems, and displays. Even some form of data loggingfharddisk) couldbeincluded. From the outset I had to be aware of both power conversion efficiency

uO U T

< “IN

a) “Buck” Step-down Converter DI

= ‘IN -

Q “OUT

VI

Cl

Ql 1 T

i 7

“ O U T ’ “IN

b) “Boost” Step-up Converter Figure 1 -Traditional DC-to-DC converter designs.

than 100 mA) converters in previous projects, I perceived this as a weekend project where I merely extrapolated and expanded on basic design theory. Now, after successfully doing it, I can tell you that while the theory indeed holds true, the proof is more elusive. General-Purpose Design Objective

Photo 2- LT 7070-based flyback converter. and power consumption. Rather than tailor a custom supply for each node, I decided to build a general-purpose converter with three output voltages (+5V, +lZV, and -12V) that could supply high currents where needed yet still have a low quiescent current when less consumption was required. Unfortunately, getting from here to there is more easy to understand as a task than an accomplishment. Since I have built many low-current (less

“,N

oL%+o -

UOUT

“ O U T - -“IN

c) “Buck/Boost’ Polarity-Inverting Converter I’N

Dl

6 01

i T

“OUT

-Nlr

YN

d) ‘Flyback’ Transformer-Coupled Converter

My design objective was to build a DC-to-DC converter that could be used as a general-purpose 12V-powered UPS for embedded systems. Its modest specification would be: +5V at 1.5 amps, +12V at 0.5 amp, and -12V at 0.25 amp. Its efficiency should let the battery last a reasonable time. More importantly, it should be efficient enough so that the current requirements of the 12V common supply are not excessive when powering multiple converters. This article documents the progression of events leading to a final power supply configuration. At the same time it answers basic questions on switching regulator-based DC-toDC converter design. For the record, I don’t claim to be an authority on this subject. This project is presented as a collection of tested circuits with useful observations because, in my experience, successful high-current DC-toDC converter designs have more to do with layout technique and analog black magic than anything as tangible as component specifications or schematics. While my final converter exceeds the design objectives and is relatively easy to build, successful duplication of it will have a lot to do with your ability to hold the magic wand properly. Since we have to start someplace, understandingthedifferencebetween plain-vanilla series-pass regulators and switching regulators provides a good introduction. I’m sure someone has already asked why we don’t just regulate the 12V common down to 5V through a three-terminal regulator. Series Pass vs. Switching Regulators Since the advent of the three-terminal voltage regulator, it seems that

everyone has become a power supply expert. No longer are ten pages of calculations required to produce a design for even a modest power supply. Three-terminal regulators like the LM317 are so easy to use that few experimenters stop to consider how inefficient they are. Consider for example, using an LM317 to power one of the controller nodes above. Given the maximum level of the common input voltage Win) for a 5-volt 1.5~amp output, 11 volts would be dropped across the regulator (in actuality, Vin will vary between 1OV and 17V depending upon whether it is operating only from the battery or from an ACpowered charger and power supply). Power is dissipated by the regulator in an amount equivalent to the differencebetween theregulator’sinputand output voltages multiplied by the current through it. The LM317 and similar linear regulatorssuchas the 7805 and LM340 are all called series dissipative regulators. They function in a linear mode, simulating a variable resistance between the input voltage source and the load. The regulator maintains a constant output voltage by dissipating the excess power as heat. Unfortunately, as we see in this example, it consumes 16.5 watts producing the desired 7.5-watt output (31% efficiency). In most applications the ease of use and relative low cost of linear series regulators far outweigh the inherent lack of efficiency. The linear series regulator is well-suited for mediumcurrent applications or applications with a small input/output voltage differential. When electricity isn’t supplied by a battery and costs only ten cents per kilowatt-hour, it’s hard to get concerned about losing 16.5 watts. Why Use a Switching Regulator? Power supply efficiency usually isn’t important unless size, heat dissipation, or total powerconsumption is limited. Since the common power source for our network controllers is a battery (when the AC power fails), we have to be more careful about how April/May 1989

11

much energy is converted for useful work and how much is thrown away as heat. Efficiency is really the name of the game. In a series dissipative regulator, conversion efficiency is directly related to the input/output voltage differential. As thedifferencebetween the two voltages increases, efficiency decreases. It would be far better if the regulator consumed no power and if all the power were channeled to the load. Of course, perfect conversion efficiency is impossible, but the inherent fault in using series dissipative regulators is the linear operating mode of the series-pass transistor. If the transistor is used as a switch (in saturated operation) rather than as a variable resistor (in linear model the pass transistor consumes very little power. (This is not a new discovery.) A regulator constructed to operate in this manner is called a series switching regulator. The same seriespass transistor switches between cutoff and saturation at a high frequency, producingasquare waveofamplitude Vin. This waveform is then filtered through a low-pass LC (inductance/ capacitance) filter, producing an average DC output voltage (Vout) proportional to the pulse width and frequency. The efficiency of such a regulator is generally independent of the input/output voltagedifferential and can approach 95% in good designs. Switchingregulatorscomeinvarious circuit configurations, a few of which are the flyback, buck, boost, and buck-boost types. Also, unlike the typical three-terminal dissipative regulator, the switching regulator can be directly configured to operate in any of three modes: step down, step up, or polarity inverting. Switching-Regulator Basics Figures la through Id outline the three common modes of switchingregulator operation. Basically, the switchingregulatorconsistsofapower source which supplies a voltage Vin, a “switch” Ql, and an LC filter. The way the components are connected determines the output mode. 12

ClRCUlT CELLAR INK

gure Z-The 78S4U

regulofor.

IS

a popular vanot3le-trequency-type switcning

Buck Regulator In the step-down buck regulator in Figure la, the basic circuit operation is to close transistor switch Ql for a time Ton, and then open it for time Toff. The total, Ton + Tofl, is called the switching period 7’. Neglecting the saturation voltage of Ql (Vsaf) and the diode(Vdiode), thevoltageat theinput to the inductor is +Vin during the time Ton and zero during Tofi (These other voltage drops would be included in calculations that choose actual components.) When Ql is closed, a step increase in voltage is applied to the inductor coil, which has the value L. However, current flowing through an inductor cannot change instantaneously; instead it increases linearly according to thefactorUdi/&J, buildingamagnetic field. This reduces any instantaneous current changeseenby theload. When Ql opens, the magnetic field in the inductor decays linearly, supplying power to the load. The current path is completed through the forward-biased flyback diode Dl. As you can see mathematically, the output voltage of a buck converter is always less that its input: Duty Cycle (DC 1 =

Vout = Vin x DC

Ton Ton+Toff

In this type of switching regulator, the inductor and capacitor form a low-pass filter. High-frequency pulses are applied to the input, and an averaged DC level comes out. The peakto-peak ripple voltage is a function of the switching period T and the values of the inductance L and capacitance C. As the frequency of operation is increased, the voltage ripple is reduced, but the supply becomes less efficient. Boost Regulator Figure lbillustratesthecircuitconfiguration of the basic step-up boost regulator. In this type of regulator, closing Ql during Ton charges the

CO OR IM GEWI E ZIP Software now allows both the original serial DTOl and ImageWise/ PC to digitize and display full-colorpictures on an IBM PC with VGA. See page 44for more details,

inductor. When Ql is opened, the inductor discharges through Dl into the load. The output voltage is:

given by:

Vout = Vin l-DC Current drawn from the input is delivered in pulses to the output load at significantly higher currents than the average load current. Both Ql and Dl must be sized to handle these increased currents. You’ll also note from the equation that in all cases, Vout is greater than Vim

Figure lc shows a polarity-invertotin

Note that the magnitude of Vout can be either greater or less than Vin.

Variable- vs. Fixed-Frequency Regulation

Flyback Regulator

Generally speaking, in all four cases, the output voltage Vout is regulated by controlling the ratio of Ton/T, which can be altered in a number of ways depending upon the control method. Two of the more common approaches are variable pulse width (pulse-width modulation) and variable frequency. In a pulse-widthmodulated switching regulator, the switching period T is fixed and the “on” time Ton varied. Conversely, in a variable-frequency regulator, Ton is fixed and the “off” time Toffis varied. Thevariable-frequency switching regulator is generally easier to design and build since the magnetic flux developed in the inductor coil during the fixed on-time determines the amount of power deliverable to the load. This eases the design of the inductor because the inductor’s operating region within its characteristic curve is precisely defined. Operating frequency, which increases proportionally withtheload,ismostlyafunction of the inductance L, capacitance C, and voltages Vin and Vout. The fixed-frequency pulse-widthmodulated switching regulator varies the duty cycle to change the average power delivered to the load. This method is particularly advantageous for systems employing transformercoupled output stages and is often used in commercial switching supplies with multiple outputs. It is more complex and uses more components than variable-frequency supplies, but theadvantagesoutweightheextracost in high-current applications. Typical operating frequencies of switching regulators range from 10 to 50 kHz. However, there are some tradeoffs. High frequencies reduce the ripple voltage at a price of decreased efficiency and increased radiated electrical noise. If the frequency is lowered, greater efficiency and less electrical noise will result, but larger coils and capacitors are needed. Also,

A final configuration worthy of interest is theflybackconverter, shown in Figure Id. Flyback regulators use a transformer, as opposed to a simple choke, to convert Vin to Vout. During Ton, energy builds up in the core due to increasing current in the primary winding. At this time diode Dl is reversebiased. WhenQl opensduring Tofi, the total stored energy is trans-

Buck-Boost Regulator

.I

Vout = -Vin x (DC /l-DC 1

be stored in the form of DC current in the windings compared to using pure AC waveforms.

lw

UIN I@-17u

0

1

T

T

Ql

1

TIP42A

1

UOUT +5u

2N4398

1

3.5A

I

.2K

lout 1

1.5A

1

Figure 3-A 7.5amp, 5-volt buck regulator using the 78S40.

ing buck-boost switching regulator. As in the other cases, closing Ql charges the inductor during Ton. When Ql is opened during To/$ there is a “kickback” voltage produced by the inductor as it discharges. This effectoccurselsewhere, too. For years, many of you have probably been putting reverse-biased diodes across relay coils, perhaps without thinking about it. The purpose of the diode is to dampen the high-voltage spike produced after a pulse is applied to the inductive relay coil. In a switching power supply, rather than short out the voltage, diode Dl directs this opposite-polarity voltage to the load. Buck-boost regulators have an output

ferred to the secondary winding and current is delivered to the load. The primary-to-secondary turns ratio (IV) affects Vout and should be set for optimum power transfer: Vout = Vin x N x (DC/l-DC) The greatest advantage of flyback regulators is that they can have an output voltage that is higher or lower than theinputvoltageandcaninclude multiple windings on the transformer secondary to create other isolated voltages. Unfortunately, all this does not come without a price. Flyback converters have higher ripple currents due to the high energies which must

April/May 1989

13

inductor and capacitor are required to make a highly efficient switching power supply. (The internal Darlington-configured transistor switch and diode of the 78S40 are capable of handling 1.5 amps at 40 volts, but an external transistor and diode are

connected to a fixed 1.3V reference voltage. If the output voltage exceeds the reference, the regulator will begin to skip Ton cycles until the voltage lowers. Changing the output voltage setpoint in a buck converter is simply a matter of changing

higher or lower output current depending upon the speed and current ratings of these devices. Switching regulators use special Schottky Barrier Rectifiers specifically for their low forward-voltage drop (0.3-0.6 volts typically) and high speed. Using a TIP42A(6-amp) transistor and lN5822 (3-amp) Schottky diode, I was able to obtain a 1.5-amp output current for a TIP32 IN5288 10=ui Vin range of 9-17 volts. By changing Y 0 UOUT R the transistor to a 2N4398 (30-amp) MILLER -12u e20mn l 4622 and the diode to an SB840 or SR802 (8amp), the output current could be increased to 3.5 amps (the current-sensing resistance was lowered to 0.033 ohms to handle the higher currents). Figure 4 outlines the mathematics involved in making buck, boost, and buck-boost converters with the 78S40 switching regulator. For themost part I used these calculations to form the Figure 5-A buck-boost design using the 78S40 generates - 72V from a fundamental basis of the end result with a certain sprinkling of empirical IO- 7 7V input. modification. Because the input voltage Win ranges from 10 to 17 volts) is MILLER l 5506 MILLER r5502 not a constant, there is actually a range 158 UH SR504 I0 -UH bd s=iii VIN of component values which are all 0 UOUT 9-IlU +12u 81.5fl optimum at a specific combination of Vin and load current. The final component values are compromise selections derived by building the circuit 78840 and testing it. TIP32 ‘.c+ E 3 An important fact about the buck CT GNO Cd %I 12 il 10 converter also pertains to boost, buckboost, and other configurations. Traditional switching regulators are electrically very noisy. Switching transients can be coupled throughout a Figure 6-Again using a 78S40, a boostdesign is used to generate + 72V power supply either inductively befrom a 9- 7 7 V input. tween adjacent components or directly needed here because of the increased the resistor divider between this through inadequate or misrouted currents involved in this design.) grounding. Grounding in a switching comparator and the output. The current-sensing resistors are power supply, like EM1 reduction, is S-Volt Buck Regulator intended to protect the switching tran- one of those black magic areas. Much of the noise generated consistor and diode rather than the load. Figure 3 outlines a 1.5-amp, 5-volt The 78S40 will stop functioning when sists of lOO-200-ns spikes which occur buck regulator using the 78S40. Oper- the voltage between sense points (pins when transistor Ql is either turned on ating frequency is set by the capacitor 13 and 14Jexceeds0.33volts. Fora 1.5- or off. It is not unusual to see 3-volt at pin 12 (usually between 0.01 and amp buck regulator, the peak current spikes (at about 30 kHz) on a 5-volt 0.001 l.t.F). Vin is connected through is typically 3 amps and a O.l-ohm output line if you aren’t careful! Eventhe current-sensing resistor to the tran- resistor (or two 0.22-ohm resistors in tually, you will discover that if you resistor switch, and at Ton it charges the parallel) is used. move the 12” ground extension jumper inductor. During Toffi the diode conAs you might have guessed, the from the scope probe,and use just the ducts and the energy in the inductor is most important elements in switching short pigtail ground across the load transferred to the load. The voltage at regulators are the transistor and the that the noise is actually less than a the load is fed back to a comparator. diode. For the same inductor value, volt. That’s better, but not great. The other side of the comparator is this regulator configuration can have In most cases, switching regulaApril/May I989

15

tor designs are greatly improved by post-regulator filtering. Capacitors alone, regardless of their size, usually don’t eliminate these high-frequency spikes. Instead, an LC filter consisting of a 7-lO+H choke and a 10004700PF capaci tor works well in most cases. Be advised that these aren’t just any old inductors. Typical low-cost molded inductors are only rated at a few hundred milliamps. Since we are talking amperes here, make sure that the filter components will handle the current. The 9-amp Miller chokes that I used reduced the electrical noise to about 200 mV peak-to-peak.

VU :

SHITCH OUT 5 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ................................................................. ___,..._____,,____.____ / 16V

FB

2; ;

Negative 12-Volt Buck-Boost Regulator Creating -12 volts using the 78S40 switching regulator is described schematically in Figure 5. The major difference between it and the straight buck regulator is that the inductor and diode are now in the reverse positions. The inductor is charged during Ton again, but this time during 7’ofithe negative-polarity EMFgener-

- - 0.15V

Figure 7-The LT7070 integrated switching regulator is often used in f,Yback co,.,verter desjgns ated by the collapsing magnetic field in the inductor is directed through the Schottky diode to the output capacitor and load. Regulation set-point

*

Your

Move.

) HIGH

MODEL PCSS-SX MULTD’ORT SERIAL ROARD

MODEL 9000 {EHE)PROM MPU PROGRAMMER

Fastest programmer on the mafket. l Quick&Intelligent programming atgorithms. * Supports megabit eproms. l Programs largest variety of chips.

) TIME SAVING \ I

_;;

MODEL ROMX-2XL EPROM EMULATOR

_II

Emulates 2716-27010 eproms. 256K to 1024K memory available. * Battery backed up, auto emulate on power-up. * Low-cost, pays for itself on first project. l Free 19.2K Serial Communications Software_

l

8 RS232 ports par card (opt. 4 port).

l

Alf 8 ports WO% DOS compatible,

* 32 ports may be added to a PG.

l Uptionat RS422 to 4WO ft. * Interrupt driven EGOS Enhancement Software included,

VematMy, High Performance and Time Savingproducts are what you’re looking for, make that move to&y and let GTEP put one, two, three or allfour of tkeseproducts to workfor you! Remember, other development products are available, -CAL&I! Iflntelligence~

l

l

16

MODEL PCSS-81 OCTAL SERLhL CO~NICATi~N C O - P R - R

8 Co-processor driven RS232 channel&card * StOS Enhancement Software included. * Elfmtnates host processor overhead. * 32.204SKByte buffer memory. l May be used standalone in custom appl.

PERFORMANCE

l

feedback, current limiting, and LC output filtering are handled much as before except this circuit is designed to supply only about 200 mA.

DEVELOPMENT HARDWARE &SOFTWARE P.O. Box mo; Bay St. LOui8. MS sszl-2310 U.S.A. OROER TOLL FREE lg002ii&OTEK (4@36f SAX: 1.Bf~1-4S7.0~6 MS b Tachnic~ suppoft 1.601487-fMm

Circle No. 12 1 on Reader Service Card CIRCUIT CELLAR INK

12-Volt Regulator: Boost or Buck?

The results of my empirical cutting-and-pasting are in Photo 1.

When I got to the 12-volt section of the power supply, the configuration was not as straightforward as theother sections. In fact, it was almost easier to contemplate going back to a lowdropout S-terminal series-pass regulator than to build what seemed to be required here. The trouble with producing12voltsfromalOV-17Vsource is that it can’t easily be done with a single-stage design. Let me explain. Using a buck-converter design we can produce 12 volts from an applied voltage of approximately 13V to 45V (theupper limit of the 78S40). Similarily, using a boost-converter design we can produce 12 volts from an applied voltage of approximately 5V to 12V (a typical12-voltoutputboost-converter circuit is illustrated in Figure 6). Unfortunately, since our input supply voltage can range from 1OV to 17V, no one circuit adequately fits the bill. Just for the sake of experimentation, I did build a buck-buck-boost compound regulator just to see if the results were worth the effort. Using a circuit similar to Figure 3, I made a 3.5amp, S-volt output buck regulator. Then, with a circuit similar to Figure 5, Iconfiguredan8-volt-to-12-voltboost regulator. Surprisingly, the compound configuration worked, but it seemedlessefficient than1 would have liked and it produced more electrical noise than theothercircuits; undoubtedly because of greater PC board real estate and radiating components.

A Flyback Regulator to the Rescue As the others around the office here can verify, I had considered forgetting this whole project idea because I hadn’t arrived at a “neat,” cost-effectivesolution to theproblem. ThePhoto 1 prototype, while workable, was hardly some thing I cared to make more than one of. It was at that point that I came across the Linear Technology Corporation LT1070 integrated switching regulator and its suggested use in flyback converter designs. The LT1070, block diagrammed in Figure 7, is a fixed-frequency current-mode switching regulator. It operates from 3V to 60V and can be used to produce the same buck, boost, and buck-boost circuits previously described as well as a flyback converter. The major advantage of the LT1070, unlike the 78S40 used at this current, is that both a 5-amp transistor switch and a 0.02-ohm current-sensing resistor are internal to the LT1070 chip. This not only reduces heatsink and board space requirements but also eliminates theradiatednoisefromconnections between these components. The shorter the wires in a switching regulator design the less electrical noise. The flyback converter circuit is very straightforward: The LT1070 closestheswitchontheprimarywinding causing the transformer core to

l

store energy in its magnetic field. A resistor-capacitor-diode “snubber” network isinserted across the primary to reduce switching transients. No current flows in the secondary windings because the diodes are reversebiased at this time. When the switch is turned off, the magnetic field collapses and induces a voltage into the secondary windings. Given the reversepolarity of the collapsing field, the diodes become forward-biased and the energy is transferred to the outputs. (Note: It is important not to confuse this flyback pulse transformer with a sinusoidal AC transformer half-wave converter design. In all cases the secondary windings use Schottky diodes and the energy is transferred in squarewave pulses, much the same way as the previously described buck converters. Use of “standard” silicon rectifiers will cause excessive power dissipation and low output voltage.) The PE-65108 pulse transformer has two primary and three secondary windings. Given the turns ratio, this circuit can work at either 6 volts with a parallel-wired primary, or at 12 volts withaseries-wiredprimaryasshown. There are three secondary windings. Two can be used for 12-volt outputs while the third is designed for 5 volts. The system regulation is controlled by feedback from the 5-volt supply section. The 5-volt secondary output usesa resistor divider network to compare the 5-volt output to a 1.24volt reference in the LT1070. If the output voltagestarts todrop, thepulse

PE-65108 INS822 0 UOUT +12u 8.733

USN +12u

-12u 8.2%

SB840 OR 662 +_3200 VFD

0 UOUT +su 62.M

Figure a- The final LTl070-based design uses fewer components than the equivalent 78S4Gbased design. April/May 1989

17

RS-232C INTERFACE AND MONITORING EQUIPMENT CATALOG

width-modulated switching regulator just lengthens Ton a bit for each cycle to compensate and vice versa. With the 5-volt secondary regulating properly (and not overloaded) approximately 14V-16V is induced in each 12-volt secondary winding. A pair of special low-dropout-voltage three-terminal linear regulators are used to create a regulated 12-volt output. LM2940-12 l-amp, low-drop out regulators require only 0.6 volts across them (Vout-Vin) instead of the 3 volts typical of 7805-type devices. The LM2940-12 is only available as a positive voltage regulator, but with isolated windings,each secondary can be configured as an isolated 1Zvolt output. One is simply inverted to appear as a negative 12-volt output. The Proof is in the Pudding

WRITE or CALL for YOUR FREE B&B ELECTRONICS CATALOG TODAY!

Pages and pages of photographs and illustrated, descriptive text for B&B’s complete line of RS-232 converters, RS-422 converters, current loop converters, adapters, break-out boxes, data switches, data splitters, short haul modems, surge protectors, and much, much more. Most products meet FCC15J. Your RS-232 needs for quality, service and competitive prices will be more than met by B&B ELEC TRONICS. Manufacturer to you; no middleman! Money-backguarantee!Same-dayshipment!Oneyear warranty on products! Technical supoti is read& available. Terms: ha, MC, cash aders pxkpaid. P.O.’ s tan q&led lrms accepad. IL reiden& add 6 i/4% sales tax. 0

rated

RUSH MY NEW FREE CATALOG

Name Company Street City

State -zip-

B&B

elertrwnirf5 MANUFICTURING COMPAN”

4032C Baker Road

l P.O. 1040 Ottawa, IL 61350

Phone: 81 S-434-0846

:ircle No. 105 on Readers Service Cart 18

ClRCUlT CELLAR INK

One final concemisradiated EMI. To minimize this electrical noise and inductive pickup, the divider resistors and the components between pin 1 and pin 3 should be mounted very close to the LT1070. A single-point grounding system should be employed with the wires routed for least radiation (black magic again). All three secondary outputs have post-regulatorLCfilterstoeliminatespikes. You’ll note that the divider feedback network is connected on the “quiet“ side of the LC filter so that output noise does not unduly influence regulator stability. This flyback regulator seemed to work quite well and noise was generally less that 200 mV peak-to-peak on any output on my prototype. I would expect it to improve on a production PC board. With a 12-volt input I was able to obtain 2.5 amps at +5 volts, 0.75 amp at +12 volts, and about 0.25 amp at -12 volts. If I raised the input voltage to 17 volts, more output current was available from all three outputs proportionally. Conversely, at a given input voltage, the 5-volt output current could be raised if the 1Zvolt output currents were lowered and vice versa. There seemed to be a minimum 5-volt current necessary for regulation, however, and a lOO-ohm resistor was added to make sure it was always

there (and should be removed in a fixed-use application). As the input voltage was lowered to 10 volts, the available output current wasalsoreduced. Conservatively, it was still 1.5 amps at +5V and a 100 mA at +12V. I tested this to the extreme and found that the board still regulated down to about a 6.5-volt input (of course if I ever had to depend on that, I’d probably have worseproblems elsewhere). If you use this design for currents lower that I specified, then you may want to use an alternativeregulator. BoththeLT1071 and LT1072 have lower current ratings and costs. The completed power supply circuit, shown in Figure 8 and Photo 2, used considerably fewer components and less board space than the previous three designs to produce the same relative power output.

Ll Sources J.W. Miller Division Bell Industries

19070 Reyes Ave.

Rancho Dominguez, CA 90224

(2 13) 537-5200

linear Technology 1630 McCarthy Blvd. Milpitas. CA 95035 (408) 432-l 900 Pulse Engineering, Inc. P.O. Box 12235 7250 Convoy Court San Diego, CA 92112 (619) 268-2400

IRS

201 Very Useful 202 Moderately Useful 203 Not Useful

Circuit Cellar Books

Circuit Cellar INK writers often refer to previous Ciarcia’s Circuit Cellar articles. These past articles are available in book form from Circuit Cellar Inc., 4 Park St., Suite 12, Vernon, CT 06066. Ciarcia’s Circuit Cellar Volume I covers artides in BYTE from September 1977through November 1978. Volume II covers December 1978 through June 1980. Volume Ill covers Juiy 1980 through December 1981, Volume IV covers January 1982 through June 1983. Volume V covers July 1983 through December 1984. Volume VI covers January 1985 through June 1986.

Ctrl

‘ls’ BLE ’ N K Tell me Why

Letters to the INK Research Staff Answers I# Clear and SimpZe

That schemafic showing the combinafions of 8031 systems was essentially correct, but you have to take the colored backI want to thank you again for the fine work you are grounds with a grain of salt. Basically, some of the gates needed doing to disseminate electronic leading-edge technology for various “subsysfems” wound up in the wrong colors simply and encouraging us to look deeply into this field and its because gerrymandering in all the peninsulas and tributaries mushrooming future. would huve complicated things a little too much. What it boils I have a few questions for you: down fo is that, if a circuit looks like it needs a connection, if does! 1. Concerning HAL-4, why are the signals to IC8 There were a few minor errors in the schematics; if you’re (74HC373) scrambled on input and then descrambled on particularly interested in the details you should sign onto the output? Circuit Cellar BBS and get fhe latest information. 2. The original schematics show the IC4 (8255) CS The trick to having both local and shared memory is that you hooked over into IC7 (74LS138), which supposedly is not must have buffers and gates to provide separate dafa paths for there until the full system is defined. The same situation each CPU, with separate address logic on both sides. If you take exists for IC5. Why is this? How should it be hooked up? alookaf thelmageWiseschematics,you’llseehow fhevideodafa 3. At a previous job, I worked on a parallel processing bus (going fo Ihe ADC and DAC chips) is separated from the system which had three CPUs running in parallel, and 8031 data bus (going to the EPROM and address latch). Trace used a memory scheme which included “shared” RAM, down the addresses though the mu1 tiplexers to see how the 8031 which could be accessed as read/write by any or all of the can bedoingsomefhingwhilefheZUMsarehandlingvideodafa. CPUs, as well as “private” RAM which could only be There are several ways to convert from CMOS to TTL, accessed by its associated CPU. I was wondering how one rangingfrom ZCs dedicated to the job all fhe way down to balingwould go about setting up a circuit layout so as to be able wire circuits you build on thefly. If depends on how many lines to do this with microprocessors. you need to convert and how many units you‘re building;a one4. I am having a bit of difficulty finding information as ofi project can use circuitry that makes no sense in a real to what to do to establish a CMOS/TTL interface. Can you production unit. give me some hints here? Rather than give you the answer, we’ll describe the process. 5. Where can one get information as to what sort of There are two conditions you have to cover: a CMOS low level current demands different devices will incur, so as to be must draw enough currenf from the 7TL gate to ensure that it able to total up a system’s power usage? “sees” a zero, while a CMOS high level must supply enough 6. What is a good way of interrogating a battery-sup- currenfforthe7TLbackleakagefoshuf offthegafe. Therequisite plied power setup so as to detect a “battery low” condi- currents and voltages differ for the various flavors of TTL and tion? CMOS logic, as well as for individual gates within each family, so there are no hard-and-fast rules. Allan R. Summers One rule of thumb is that you should nor have a singlegate Pasadena, TX driving both CMOS and TTL loads; stick with one or the ofher. This simplifies building the converters because you don’t have fo worry about the effects of drawing nonstandard currents from Dear Allan, thegafe on the rest of the system. The data books that give you enough informafion to build Thanks for the kind words on OUT projects! We t y to strike level converters also tell you about the maximum and average a balance between “leading edge” and “practical” projects; current requirements for the complete chip. Mail order parts letters like yours tell us that we are on the right track. sources usually advertise the National and TZ data books; take a The scrambled lines on Hal’s ZC8 make sense if you look at lookat the Digi-Keyand Jameco catalogs for thespecifics. Make the printed circuit board. Assigning the bits in that order made sure you gef the Application Nofes books, too, because they have the tracesfit on a double-sided board! Since this depends on fhe lofs of recommendafions and hints for handy circuits. particular board, you’ll find differenf layouts on all the projects; As you might expect, defecting a low-battery condition ve y often we leave the choice of bits up to the guy who does the requires an analog-to-digital converter. Forfunately, you only PC board bayou f. need one bit, so it isn’t too hard to build, sincesomething likean 20

CKCUT CELLAR INK

LM313 comparator that compares a referenceagainst the battery voltage fills the bill. When the battery falls below the reference voltage, the EM312 output changes state (either high to low or low to high, depending on your circuit y) and your CPUsfarfs saving things. The frickis that you’vegof to beable to correlate the battery voltage with remaining charge. Thevolfagevaries with ambient temperature, as does the available charge, so this isn’t quite as simpleas if sounds. Wedon’f recall fheparf numberofiand, but we think National Semiconductor makes a battery-voltage monitor IC that handles some of thegrisly details. Once you get fhedafa books you can fakea walkthrough the “building blocks” section and find something useful. You cangef rapid answers to questions like these by signing onto the Circuit Cellar BBS. There are a bunch of really competent folks hanging out there who can probably answer any question you can dream up. Best of all, the whole group can suggest approaches you haven’t thought of and describe all the details based on real experience. A Problem of Power I have a car restoration shop. My problem is that my CAD system (AT clone running at 10 MHz) sends out beeps and stops when we use our plasma cutter, which has no HF to start its arc. Our telephone system, a genuine Bell setup, lights up all six buttons and won’t ring when our HF piggyback arc stabilizer runs when we’re using the heliarc welder. Is the fix as simple as adding ground wires? Or should I build a screen room around the welder/cutter? How about this as a project for your magazineYZlean Power in an Ugly World”? Robert J. Schumann Kansas City, MO Dear Robert, Every now and again wegef a letter that reminds us of just how odd things get out in the real world.. . The hash in your AT and phone system is probably coming through the power lines. If could be radiated (we bet you don’t have any background music playing while you’re welding, do you?), but we think the place to start is with the line cords. Any electric arc will generate energy across the entire electromagnetic spectrum, quite literallyfrom DC to daylight in fhecaseofa welder. Nafurallyenough,fhemorepowerinfhearc, the more power shows up as interference. What you need is a filter foremovefheobjecfionablefrequencies whileleffingfhearc burn normally. From our experience with EMI (electromagnetic inferferencejgenerafed bycompufersysfems,aferrifefilferis the way to go. Ferrite filters are made up ofafinely divided iron compound with a high resistance to electrical current. A current-carrying wire passing through a ferrite slug induces a current in the

ferrite, but theferrite’s resistancedissipates fheenergyveywell. The resistance increases withfrequency, so fheslugforms a quite effective low-pass filter. One catch is that the ferrite slug must be able to handle the induced current without saturating. In the case of your welders and cutters, that’s going to fake a pretty big chunk! Rather than paying real money for this project, fake a trip to the local junk yard and scavenge the yokes from a couple of TVsefs-the older, the better. If you’ve never rummaged around in a TV bt$ore, what you’re looking for is the deflection hardware around the neck of the picture tube. Along with all the coils is a big hunk of blackferrite, which is just what you want. You may have to break the tube fogef the yokeoff, but that’s why you do this trick in the junk yard instead of Sears. Do be careful, though, because the implosion resultingfrom shattering thepicture tube can blow glass all over the neighborhood. Wear a face shield and gloves. The safest method is to put the neck of the tube in a plastic bag and rap if with a screwdriver. This presumes, of course, that the junk yard has no further interest in the tube! With a few yokes in hand, simply string the power lines through them. If the power lines go directly to a junction box on the wall, run the cables on the secondary side through the yokes. The key point is that the ferrite slug should form a continuous path around the current carrying wire, with one wire per yoke. Don’t bother wrapping the wire around the yoke, because one “turn” is enough; more turns simply builds a step-down transformer with a single-turn shorted seconda y. You should see an immediate improvement, but if not, f y moving the yokes closer to the workpiece. Because fhearcs are the source of the hash, the less wire carrying the current, the less interference will be induced elsewhere in your building. As an alternative, fy puffing a yoke or two on the power lines leading to the AT and phone system. After all, if doesn’t matter where you filter the hash out as long as if doesn’t get to thecircuit y. You will have fofilferall of fhepower lines leading to the AT; don’t forget the printer and modem! We’ve been thinking of doing an article on RFI control and your letter has pushed us over the edge. It’ll probably show up around the end of 1989, simply because we’vegof so much other stuff to do between now and then.

IRS

204 Very Useful 205 Moderately Useful 206 Not Useful

In Visible INK, the Circuit Cellar Research Staff answers microcomputing questions from the readership. The representative questions are published each month as space permits. Send your inquiries to: INK Research Staff c/o Circuit Cellar INK B o x 7 7 2 Vernon, CT 06066 All letters and photos become the property of CCINK and cannot be returned. April/May 1989

21

PRODUCT REVIEWS:

The Next Generation

Circuit Cellar INK sells out and enjoys it!

t’snot often that one is witness to a major event in human history, but if you’re reading this, you’re not only a witness, but an eyewitness! Yes, you see before you an occurrence of epoch-changing proportions. Mere human inventions pale in significance before this. Polio vaccines, atomic energy production, and those little plastic things that slowly slither down your walls all fade away when compared to what you hold before you. What am I talking about? Why the Circuit Cellar INK Benchmarking, Product Evaluation, and Junk-Food Consumption Testing Facility. Now, I know we said that you wouldn’t see any “metoo” reviews or evaluations in Circuit Cellar INK, but hey, times change. First, we saw what was being passed off as product reviews by other magazines and realized that no publication in this end of the universe was better qualified to write technical, hard-hitting reviews than were we. Second, we were getting miffed that companies weren’t sending neat toys (free of charge) to our offices. Third, it seemed like a good way to get paid for breaking stuff.

I

No Dweebs Here At Circuit Cellar INK, when we decide to do something, we don’t go in for half-measures. Before we opened the CCINK BPE&JFCTF, we performed an exhaustive study of what “the other guys” were doing. What did we find? Wimpy softwarebenchmarks and effete instrumentation, that’s what! Where were the calisthenics for individual instructions, the particle accelerators, the blast craters? They were not to be found, at least not until now. We have assembled the only testing facility of its type in the free world. On the hardware side, we put together a state-of-the-art laboratory filled with expensive gadgets, heavy industrial machinery, and scary electronic stuff. For wringing out the software, we contracted with a littleknown hacker/genius to write assembly code that’s so compactly written, so immune to optimization, and so thorough in its exercising of undocumented features, that we’re not sure what the hell it really does. Finally, and most importantly, we’ve assembled the strangest group of gonzo testing weenies ever to don lab coats. These guys live to trash computers, and subsist on Chinese take-out, Cheeetos, and Jolt soda. Is our crew qualified, or what? 22

ClRCfJlT CELfAR INK

let the Games Begin As we began looking through the literature on evaluating systems, we worked our way through the Whetstones, Dhrystones, Savages, Livermore Loops, and many more. We found that none of them really told volume purchase influencers what they needed to know about the system at hand. We thought about presenting all of the tests in an easy-toread chart format so that readerscould form their own opinions based on the information presented. Then we came to our senses and realized that what readers really want is to be led by the nose, bludgeoned by a single sledgehammer factoid that they can point to in self-defense when things fall apart. If it’s a flattering number that companies can use in advertising to justify their high prices, all the better. With all of the above in mind, we formed our “Magic Number Search Committee.” We knew that the most important piece of the puzzle was the name. After hours of research, pages of dictionary work, and several intense rounds of Balderdash, we found The Word. Henceforth, when systems of any type are compared, they will be compared in our devastatingly elegant unit of measure, the Mopoke. The range of Mopoke is from -53 to 177, using a logarithmic scale. Using the logarithmic scale makes for snappier graphics and often makes many mediocre products look far better than they truly are. Inside the Mopoke Where most benchmarking programs content themselves with simple performance measurements, you must

understand that the Mopoke is a unit for measuring far more than mere performance. In deciding to use the Mopoke, with all its subtle complexities, we acknowledge that most users are interested in more than how fast the processor can idle, or how many bits can be blasted to the disk per second. No, most users really want, and Circuit Cellar INK alone is now able to offer, an objective evaluation of the true gestalt of the system. We start, of course, with performance. The first stage of Mopoke evaluation is to individually exercise every single instruction possible in a given system. Each instruction is performed, with no setup, overhead, environment, or operator intervention, for one million iterations. The time for each individual iteration (timed and recorded by a special nearly noninterventionist external CCINK BPE&JFCTF clock that is linked by special optical links to the Atomic Clock at the National Bureau of Standards) is logged, and the results are averaged using a geometric mean. Next, polar Fourier transforms are performed on the total data set at S-degree intervals. These results, along with the geometric mean from the first evaluation, become fodder for the Mopoke canI non. Next, we fearlessly dive into the human factors morass. Our dedicated staff painstakingly measures keyboard feel (key travel, key cap size, depth at roll-over, angle of attack, height of the little bumps on the home keys), display ergonomics (dot size, dot pitch, glare, aspect ratio, jitter, swim, bugaloo, accessibility of controls, versatility of inputs, and how hard it is to sneak the thing home to use withyourVCR),and mainunit construction (footprint, depth, height, hat size, presence of nifty LED displays, decibel level of the fan, pitch of the fan, aesthetic considerations). We then go whereno review has gone before and quantify what can only be called “Ego Appeal Factor” (EAF). The EAF takes into account important issues such as: Will this system make your boss jealous? Does the color of the system complement your power ties? Do the noises of start-up guarantee that everyone in the office knows when you fire that sucker up? We run the results from all of these scientific tests through a weighted averaging system, where weights are determined using a complex system based on the researcher’s biorhythm, the hourly exchange rate between the U.S. dollar and the Greek drachma, and the Solunar Table published in Field and Stream.

We don’t overlook reliability in the Mopoke, either. We are the only publication that runs each and every test unit through both a steel annealing oven and a cryogenic life-extension chamber to check for continued operation at temperature extremes. We also operate each system in booth #7at the “Tans for theMemories” tanningsalonand at the bottom of a washtub filled with Evian water to test for susceptibility to damage from humidity extremes. Our line-noise isolation test includes operating the system on the same line as a heliarc welder, and injecting a signal taken from side B of Def Leppard’s most recent single into the test AC line. As thorough as are these tests, we recognized that most failures involve hard disks. To stress-check hard disks of portable computers, we bolt the little monsters to the main oscillating mixer down at “Merle’s Paint’N’Plaster World.” Desktop units are subjected to a patentpending procedure we like to call the “Extreme Prejudice Test.” Suffice it to say that this rigorous examination of Winchester technology involves expensive disk drives, high-velocity ammunition, and a very low “Pass” ratio. Now, most magazines would stop right there, but we’re committed to press on. We know that most users are far more afraid of their system’s manufacturer failing than of the system itself failing. So, in an action unprecedented in computer journalism, we rate the reliability of the manufacturer. First, we use the standard ratings based on information from Standard & Poor, Dun & Bradstreet, and the men’s room attendant at the New York Stock Exchange. Next, we test based on “look and feel” issues such as color (or presence) of the CEO’s hair, number of company executives wearing brown shoes with blue suits, glossiness of the annual report, and quantity of shrimp served at the company’s COMDEX party. Finally, we take a hard look at marketing expertise including public relations budget (did they bribe us?), advertising budget (do they advertise with us?), and use of inappropriate celebrities, water fowl, or hors d’oeuvres in their ad campaign. As with the other categories of data, company reliability information is not left to peacefully ferment. No, we run the raw numbers through a data colander unmatched in subtlety and precision. The results are factored with numbers taken from deep between the lines of the Wall Street ]ournal, New York Times, and Daily Racing Form. When we’re through, you’re presented with the one number that can indicate beyond a shadow of a doubt whether you should place your trust in a company by purchasing its product. Several Wall Street heavies have come to us begging for our numbers, but we have steadfastly refused. You see, we value the welfare of our readers far more than we value the paltry few million dollars offered for our secrets. Knowing that you will be able to make major purchasing decisions based upon what you read here is worth more to us than miles of yachts or buckets of caviar. So now you know the why and the wherefore of our review process. As you turn the page, do so with the proper reverence, for you are truly taking part in a new era: The Age of MOPOKE. 0 Aprl//May I989

23

W

ith our powerful new benchmarking tools in place, we decided that the first review should be comprehensive, a detailed look at a broad cross-section of powerful computational tools. For years, limp-wristed computer journals have been telling you that there’s no real way to compare different system types on an even and fair basis. Poppycock! With the Mopoke at our disposal, we can readily compare Apples and oranges, DECs and Wangs, and IBMs and Banana Juniors. Knowing that our readers are nothing if not diverse, we chose four very different computing solutions to compare in our first review. For the volume purchase influencer, we picked the GeneriComp 286, a plainvanilla PC/ AT clone. The more traditionally minded among you should be cheered by the inclusionof the Googolplexx 100, an S-100 mainframe system. For the aficionado of portable computing we picked the most powerful computer, in the smallest package we could find-the Hewlett-Packard HP-41CX. Finally, just to prove that there’s no place the Mopoke can’t go, we included a simple, dedicated system, known and beloved by many as an essential tool for modern living. We‘re talking about the Proctor-Silex TPB-5342 Dual-slot toaster. There you have it, the lineup for the first iteration of the most comprehensiveandimpressiveproductevaluation system in the history of mankind. Now let’s move on to the tests.. .

The Hardware

The GeneriComp 286: If you’re looking for the most risk-aversive choice in office productivity solutions, you’ll be hardpressed to find a more conservative computer. The safety-conscious ~-MHZ top speed, the haven of an Industry Standard Architecture (ISA) bus, and the soothing beige

Supercharged Worry Munchers

Circuit Cellar INK looks at four diverse applications solutions

color unite to make this computer a “safe” option in many environments. We got this baby with 1 MB RAM, EGA, one 1.2-MB floppy drive, a 40MB hard disk, lOl-key keyboard, one parallel port, and one serial port. A copy of GeneriComp MS-DOS version 4.0 came with the system, along with a handsomely photocopied Users’s Manual. If CYA is the order of the day at your office, the GeneriComp deserves strong consideration. The Googolplexx 100: In days of yore, when men were men and computers could heat the south wing of your house, Googolplexx became the computer marquee of choice among those truly “in the know.” The famed Googolplexx One has been updated through the years, with thelatest batch of revisions bringing us to theGoogolplexx 100. As with all computers of this ilk, the 100 comes out of the box pretty much useless. We stood around and gaped at the 43-slot backplane, the 382-watt power supply, and the variable-pitch 1.5-horsepower cooling

fan. When we came back to earth and realized that testing would be hard with no components, we ordered a 6MHz Z80 CPU, 1 MB of RAM (bankswitchable in 64-KB chunks), a quad serial port board, a parallel port board with cache, a DEC VT220 terminal, two 8-inch quad-density floppy disk drives, and a 40-MB hard disk. We also decided to get a five-foot-high equipment rack in laboratory blue to keep everything corralled. For software, we stuck with CP/M version 2.2. Documentation included 47 separate User’s Guides, Technical Guides, Application Guides, and Schematic Supplements. No overall index or bookshelf was included. The HP-41CX: If cranking away at numbers while you’re bouncing around in the back of a Jeep is part of your job description, then a portable computer may well be high on your wish list. We chose the HP-41CX as an example of a portable computer that makes sacrifices in user interface in order to concentrate on the portability side of the portable computer equation. The 41CX packs 8 KB of memory into a package less than l/20 the size of the GeneriComp. The HP comes with a one-line display, 39-key keyboard, and business-like dark brown case. It has a built-in clock/calendar and four proprietary expansion slots. We chose to get the handy leatherette carrying case (belt loop included) for better portability. The 41CX comes with an operating system and fully functional software. Documentation is in two useful volumes. Proctor-Silex TPB-5342: Who hasn’t experienced the warm feeling of waking up to hot chocolate and a couple of pieces of crunchy buttered toast? When we were trying to decide which system to pick to rourid out our review, this sturdy chrome beauty was the nostalgic unanimous choice. T h e h a r d w a r e inApril/May I 989

25

New! Modular Programming System FROM MODULAR CIRCUIT TECHNOLOGY This integrated system is Ideal for developers-it easily expands as vour needs crow! All the mod&s uke a corn&n host adaptor card so you need just one slot to program EPROMS, PROMS, PALS and more!

$29.95

l

A umversa mterface for all the programming

. User-selectable programmable addresses prevents addressing conflicts l Includes a high quahty molded cable MCT-MAC xu~ xv

.*>%+ 11 as